Thursday, March 7, 2024

AI & The Deconstruction of the Art of Photography

A barn on the old Beacon Coal Mine Road, just south of the Allentown neighborhood near Seattle. I took this pic long ago, with a Kodak Duaflex camera that I got for maybe $4 or $5 at a thrift store. Developed the film (110 Pan-X), and I made the print myself. Of course, this was during the film camera era -- obviously long before the 2000' s. The barn in this pic is long gone. The winding, narrow Beacon Coal Mine Road still looks like the gravel road it probably was when it served the (long gone) Beacon Coal Mine in the 1920's. It's been paved since probably the 1930's, though, but it always is a scenic, rustic foray into what the region used to look like before it got over developed.

I have been into the 'art' of photography since my college days, when a buddy of mine, Chris Styron, taught me how to frame a pic, focus the lens, use the F-stop creatively and accurately, use a 120mm telephoto or extender, "push" film, develop and print black and white photos in a darkroom, and the like.

Now, to most modern day folks who take pics with their smartphones, such terms like "F-stop", "120mm telephoto", and "pushing film" probably sound like alien lifeforms, but back in the days of the film-based SLR camera, they were standard operating terms.

Seattle area Rock band Rail (earlier known as Rail & Co.) playing at Highline College when I worked on the newspaper there. I took this with a Canon TX -- their working class model of SLR. It got the job done. The photos of this performance came out well, even without a flash, because Rail were starting to get big at the time, and they had a good lighting system. I "pushed" the film, something that was common among news and black and white photographers at the time.
Rail's lead guitarist, Rick Knotts, who was quite good. Rail started as a covers rock band, and then graduated to their own stuff. Knotts' Flying V was his mainstay guitar for years.

I was also heavily involved in the Highline College student newspaper at the time, and over my time spent working there, I contributed a lot of photos. Even after that, I stuck with photography, although it was placed on the back burner for a long time. My old Canon TX began to act up. Film got more and more expensive. Some film, like Kodak Tri-X and Plus-X (Plus-X was my favorite) were no longer available in stores. Even color slide film began to get rare to find. So I really didn't take many photos for a while.

When I got my first digital snapshot camera in 2004 -- a Sakar camera which was a Christmas gift -- it prompted me to get back into photography.

The Sakar was a 'cheap' digital camera, yes. But you could eke some decent, 640 by 480 pixel resolution (the VGA resolution that was common in the 1990's) shots out of it.

I even took some pics of a rock show on one of those $10 digital snapshot cameras in the mid to late 2000's, the band being the grunge band Sweetwater, who reunited after a several year hiatus. If I can find those pics, I'll post a couple here.

A pic of Rich Credo's Les Paul Gold Top that he played during "Head Down". Credo is (or was) the guitar player for Seattle Grunge band Sweetwater, who had a major label release in 1992 but never hit it as big as the big name Grunge era Seattle bands. They reformed in the early 2000's and a buddy of mine invited me to go see them with him when they played at the Crocodile Cafe in Belltown. Probably 2007 or so?
Adam Czeizsler, the singer for Sweetwater, when they reformed and played at Seattle's Crocodile Cafe in 2007 or maybe 2008? I took both of these pics on a little $10 Sakar digital VGA camera I got at Walmart or Walgreens. The camera was my first digital camera -- run off of one AAA battery, and the camera really was not much larger than that -- maybe two inches by one inch square. Uploading the pics to computer was done via a USB cable. It worked. 

When the mid 2010's hit, and I started this blog, I naturally upgraded. I got my Fuji snapshot camera in 2014 or 2015, and then I got my Nikon L32 when the Fuji froze up on me (the Fuji AX100 since fixed itself -- the microprocessor, which locked up when the battery got low, needed to completely bleed a charge -- I have a blog article on that).

All this said, I am not a terrific photographer, but I have some experience in the art. Many of the pics I have taken over the years, I have posted on this blog. 

This is a pic of some flowers at Highline College, which I took with a $24 Sakar camera, if memory serves. The colors were naturally saturated with some of these cheaper digital snapshot VGA cameras. I remember reading negative comments about Sakar cameras at the time. But if it weren't for the cheaper digital cameras like the Sakar ones, I wouldn't have pics of my cats, weekend vacation trips, and the like. My film camera was no longer working and I couldn't afford a good digital camera like a Fuji or Nikon at the time. So I made do.

However, one thing I have noticed over the past few years is there has been a decline in the quality of photography one sees on the internet. As more and more people use their smart phone cameras, and those cameras alone, you can see the graininess in the photographs. It's like when the digital snapshot camera began to be replaced by the IPhone in 2006 or so, the overall quality of photography dropped.

Now, I have a smartphone -- a TCL / Alcatel. I have used some of the pics here on my blog, as it's easier to just use my phone, instead of drag my Nikon L32 snapshot camera around with me. The photos are OK in quality for blog and some other uses. And they're certainly better than nothing.

But when you zoom in, that's when you see just how grainy cellphone camera pics -- even in 2024 -- really are. The "focus" is not much different from what you'd get with a Kodak Instamatic camera in the 1970's or 80's... It's single focal length, and limited focus. As for "telephoto" function on a smartphone, sometimes it looks good. Other times, it looks like crap, depending on the cellphone used. The autofocus can also play fast and loose with the clarity, depending on lighting. A computer app can only guess so much at the true detail in a 'telephoto' shot.

Randy Hansen playing a small pub on south Central Avenue in Kent, WA, in 2010. This was taken with a larger version of a Sakar camera, one with a different brand name but it came from the same factory. It was a 1 megapixel camera, the quality was not too much different from what you would get from a flip phone camera at that time. 
A pic of the 'famous' Selleck windmill. Selleck is an old lumbering town in the SE corner of King County, an area that still looks like something out of Appalachia. There used to be a bunch of lumbering, coal mining, and railroad towns in the area. By the time I was a kid, they all were ghost towns, really -- Selleck being one of them. I don't know what the windmill operated originally, but by the time I took this pic in 2009 or 2010 or so it was more decorative than anything else. I took this with the Sakar VGA camera I got as a gift on Christmas, 2004.

Digital snapshot camera shots from the 2000s were much better quality. They had larger aperture lenses, and real telephoto capability.

So, with all the tech available today, why are most of the photos similar in quality to a 1970's era Kodak Instamatic? Especially when there is so much computer power available in a smart phone's camera app?

The answer is pretty simple: DIGITAL SNAPSHOT CAMERAS HAVE REAL LENSES -- that's why.

Even with the latest versions of smartphones out there, the quality of the photographs aren't quite up to digital snapshot camera level resolution. This is because there are limitations -- physical, and practical. There is really only so much that a computer app can pull out of a tiny, 1 to 2 mm pinhole lens -- or even 2 or 3 pinhole lenses working together. The physics of lenses demands that you need real glass, and a real aperture, to get really high quality photographs.

Smartphone cameras (and to a certain extent the digital snapshot cameras that preceded them) use computer software to improve, sharpen, autofocus, and maximise whatever image, that makes it through their pinhole lens, casts onto the sensor. However, there's a point beyond which you're asking the software to sharpen, focus, improve imagery that isn't really there, because of the limitations of what basically is little more than a pinhole lens.

It's like the software uses a primitive form of AI to re-imagine what it thinks the camera pinhole lenses are 'seeing'. Most of the time, it works OK for social media posts and the like.

So, for most uses, I guess it's not really that big a deal. Considering the miniscule size of the lenses on smartphone cameras, it's amazing that they actually take decent photos at all. In a sense, it almost borders on the miraculous.

But zoom in? The photos are still fairly grainy.

The pictures work. But do they match the quality of photos from the SLR era (1970's through the 2000's)? Not really.

So, what does this say about the quality -- and resolution -- of photography in general?

Why are we celebrating the return of Kodak Brownie or Kodak Duaflex quality resolution (and believe me, I experimented with using those cameras in college, and sometimes you could get good results)? Is it a step forward, or a step backwards for photography overall?

Boots Jr. balancing on a fence in 2010. I took this with my Sakar VGA camera. If you zoom in, you can see it's a bit grainy. The sensors used in these VGA digital cameras weren't the best in the world, but they still gave you a halfway decent pic. If it weren't for my Sakar camera, I wouldn't have many pics of Boots. I got her in March 2007 and she died a couple hours into my birthday, in October of last year. 16 and a half years old.
Looking at pics like this you can sense the irony -- when you look at the resolution aspect of photography -- for all their expense and technology, modern smartphone pics are still a bit grainy.

THE DEMISE OF THE CAMERA, AS A STANDALONE DEVICE
OK, so I have made it clear that I think that relying solely on the smartphone "camera" -- although quite handy, and perhaps initiating millions of people into the idea of embracing the art of photography -- is a step backwards. It's handy, yes. At a distance, the pics look good. OK. But resolution? It's still grainy. Focus? It's marginal, but workable.

Even as late as the 1990's, you'd see people with cameras. Tourists would have cameras with them -- even during the digital snapshot camera era, which lasted from the early 1990's, to the late 2000's / early 2010's.

Today, seeing ANYONE with an actual camera is as rare as hen's teeth.

And it's doubtful that the smartphone is going to sprout a real camera lens -- by "real", I mean one that at least matches the 4.3 mm, actual polished glass lens that my Nikon L32 uses. Smartphones increasingly rely on multiple tiny lenses, and use app software to make up for what the lens and sensor do not actually see.

I don't think there is any turning back the clock, however. The smartphone -- a device that has replaced -- or will replace -- almost 100 different devices, is here to stay.

And its imprint on the nature of what we call photography is also here to stay. I don't think there is any denying that. 

I get it. I use my TCL phone's camera a lot. It's handy. The quality is marginal, but workable. It gets the job done. No one is going to zoom in on my pics to see if they're in proper focus, or there is any graininess. So, obviously, smartphone "cameras" are the dominant "cameras" being used probably by 99% of photographers today.

That said, a new wrinkle has appeared in the photography field that may completely change photography even further -- almost making the taking of photographs redundant for many, if not most purposes.

Welcome to AI.

THE ENTRANCE OF AI INTO THE PUBLIC'S CONSCIOUSNESS
Now, like many out there, I first learned about AI when various news articles mentioned it in past decades, and I think it was periodically covered on Art Bell's famous overnight radio talk show -- and Coast To Coast AM that succeeded it when Art retired.

I think AI was also mentioned in some Science Fiction novels and movies. "Hal", the famous self-aware computer in the futuristic 1969 movie 2001: A Space Odyssey, would be an example. I suppose the robotic characters in the Terminator movie series would be another example -- highly "intelligent", self aware (or programmed to be self-aware) computers, programs, and robotic devices that would be able to learn, "think", and the like.

There was a real intense, futuristic 2001 Stephen Spielberg movie about an AI, robotic boy. The movie came out when AI was more or less a futuristic concept, thought to be impractical at the time the movie came out. Remember, in 2001, the average PC had less computing power than your average second-tier smartphone has today. 

The Spielberg movie was called AI: Artificial Intelligence. The movie was a fairly realistic look at a future where, because of depopulation, robots are used in large numbers. Many of the robots were androids -- realistic looking, robot humans. Couples who couldn't have kids could purchase a AI capable robotic child, who for all intents and purposes was just like a real child. The movie looked into some of the ethics of the use of AI and robots, and also has cloning included in its final scene. 

I first saw this movie in 2008 or 2009 or so, when I found the VHS at a thrift store. At the time, the concept of AI, AI robots, cloning, and the like struck me as all too futuristic to think practical to be used in my lifetime.

Then, in the 2010's there were articles in the newspapers about AI, and how it would replace actual human workers. I remember reading several such articles in the years 2011-2014. Some experts on computer tech and futurism speculated that AI would replace 50% of the labor in the industrialised world by the year 2050. That hasn't happened yet, obviously. But, so far, AI hasn't spread to the labor marketplace as much as it may in the future. As for AI robots -- that still seems decades away, although there appear to be a handful of prototypes.

Of course, over the past 3-4 years AI has become a much bigger deal, and it seems to make the news daily. There are YouTube videos of AI versions of the Beatles, Johnny Cash, and other famous artists singing stuff they would never otherwise sing. There are big tech companies like Microsoft and Google involved in AI. Elon Musk apparently is interested in AI -- its negative aspects, as well as its more positive ones. According to some in the indie publishing field, Amazon apparently has used AI bots in many of its internal operations. There were at least a couple news articles in 2020 about Amazon firing people via AI bot.

As I've mentioned previously on here, I'm an indie author, and the prospect of AI generated stories and books replacing fiction and non-fiction writers' works has been discussed on an author's forum I go to. The consensus is that AI may eventually damage the fiction writing marketplace -- but such damage may be a few years, or a decade or more -- away. There is also the thought that fiction readers may still want to read works written by actual humans -- not dissimilar to the non-GMO and organic food movements -- people often will prefer the 'real' to the computer or tech originated 'unreal'.

Either way, AI is obviously here to stay.

So, what does this all have to do with photography?

A lot, actually.

I'll admit, I hadn't been aware of AI's capabilities in producing photographs until recently. And what I have seen is shocking. There are still some glitches in the AI generated photos here and there, but the results sometimes can be alarmingly genuine looking.

I saw some pics that were unbelievably REAL looking. They are so real that some in the stock photo field (who go to a radio forum I hang out at) say that they are concerned that AI may eliminate the need for stock photographers completely.

WHO NEEDS A CAMERA ANYMORE?
After seeing a few of the AI generated photos involved, I did a little research, and it appears that some of the larger AI photo generation websites and apps gleaned millions, if not billions, of photographs off of the internet, and they use a mixture of those photos to create, or generate, the photos they produce. Although the AI can use the actual images, it's my understanding that AI also can meld the real images into composite, but real looking ones. I.e., AI can take five or six faces, and make a composite, generated face from the five or six real ones it gleaned from the internet.

You own mind does much the same, when you dream. If you've ever dreamed about a person you never met in real life, sometimes your subconscious is drawing characteristics from any of the thousands of real people you've seen in your life, and combining faces, body characteristics, and the like. AI apparently works similarly.

So, in an AI pic, maybe you'll see a young woman holding an apple. The apple's image may have originated off of one website -- an advertisement, perhaps. Or maybe it came from a stock photo. Or maybe the apple image originated from a photo someone took and placed on their social media, or a large internet photo webpage. Maybe the AI used images of two or three similar looking apples, and combined the images. The AI may actually alter the appearance of the apple, before it uses it in the generated photo image. Maybe the instructions told the AI to take a Granny Smith apple and make it look more like a Gravenstein or Red Delicious one.

As for the young woman, her face, her body, could be an amalgamation of three or four real female images from the millions (or billions) that the AI app pulled off the web.

Looking at such a pic, one wonders how close the image of the apple is to the real, original image (or images) that the AI app used to generate the apple in the resulting pic. One wonders how close the face of the young woman is to the face (or faces?) of the real woman that the AI used to generate the image of the woman in the picture. The clothes? Maybe those images came from ads or stock images. Maybe if the AI is "smart" enough, it creates its own, based on the standard template for the words "shirt", "dress", "pants", "shoes" etc. that were programmed into the AI, to depict generic styles of clothing.

Admittedly, I really don't understand how AI actually works. All I know is the little I've read on Stable Diffusion and another AI website, and a few postings elsewhere -- and it's really hard to understand the descriptions of how AI operates. To me, trying to understand how AI generates photos is like trying to understand calculus, or nuclear physics. In other words -- aside from the absolute basics -- to me, understanding how AI works is inscrutable.

But AI images are almost scary in their realism sometimes, and I can see a day where -- if one wants to show friends a pic of you at the beach -- you just tell the AI generator to take your own pic that you uploaded, and tell the AI to place you "on the beach". Any beach. 

In fact, I could conceivably tell an AI generator to place me on Scarborough Beach in Perth, Australia -- a place I've never been -- and if the AI is good, it could make the photo look so real that you would think that I've been there.

Get the idea?

If you have an app or program with that much image generating power, WHO NEEDS A CAMERA?

And that is the problem. If you -- like me -- used to really enjoy taking photos, and taking quality ones, AI may cause the average viewer of your pics wonder which AI generator you used to produce that pic of a cat, or of your vacation at the Ocean.

You may post pics of you on vacation at Cozumel, and people will ask which AI generator you used, because, "wow, that looks pretty real -- looks like you were actually there." Fat chance of you actually going to Cozumel. Let's say you didn't actually go, for whatever reason... The AI can make it look to others as if you actually went there.

Now, most people will still take real pics with the real cameras on their real smartphones and post them on Instagram, Facebook, and other social media just fine.

But when AI becomes truly ubiquitous, all bets are off.

My cat Squeakers, a couple days after I got her in August 2010. The pic was taken on a $24 Sakar camera. You can see the cases for it and another, backup Sakar on the inntable next to the kitty. She always was attentive. Very smart. Still is today, Thank God.

And pro photographers? Will they really have much of a job? Are there many actual pro photographers today? Who hires a photographer for a wedding anymore, when everybody has a smartphone? And with AI, they can make it look like you held your wedding in the Notre Dame cathedral, or some other fantastic setting. High school graduate pics have been another realm for professional photographers that AI is probably going to obliterate. With AI, the grad can be dressed however they want, with as many clothing changes as they desire, and will be depicted in whatever location they want to be "photographed" in.

In other words.... No pro photographer with a large format, digital SLR camera will be necessary.

And I clearly don't see much of a future for digital cameras with real, operating lenses -- the snapshot cameras, the digital SLR's, and the like -- once AI becomes so powerful you can dial up whatever you want on a computer screen. For one thing, cameras aren't cheap. And the rarer they get, the more expensive they'll get. I'm certain that there will probably still be a market for some higher end cameras -- new source materials for AI generated images will still be needed. And those new source materials will be taken with actual cameras.

And there will be always a marketplace for phone camera 'selfies', and the like. But serious photography? It's probably going the way of the Dodo.

We're entering into a world where "photography" is smartphone, pinhole lens cameras and AI, with AI taking more and more of a part in the field.

I'm not saying that's necessarily bad. It is what it is. But -- like many other fields I grew up to love -- rock music, electric guitars, photography, creative writing, MW and SW radio, etc. -- the advances in technology are transforming them slowly into oblivion, or altering them completely into something ultimately different from what those arts and fields were just 30 years ago.

I can see where smartphone manufacturers will include more AI into their photo and camera apps. The AI may improve the results, making the grainy images look more accurate in nature, and improve the focus. I'm certain there is already a certain amount of computing power included in your average IPhone or Android's camera apps that would be considered a form of AI -- but that tendency is only going to increase. As AI is more completely embraced, and it becomes more reasonably priced and compact enough to include in a device like a smartphone, all bets are off.

And that won't eliminate the fact that the use of the AI in a smartphone (or other) camera will only replace the real with something generated. 

There is an old adage, taken from a 20th Century novel: "The past is a foreign country, they do things differently there." (British author L.P. Hartley, The Go-Between, 1953)

One could say the exact same thing about the future.

Here's the Seattle band Rail playing what was their first MTV hit, "Hello". They went on to become a second tier hard-rock and metal band, touring the US and selling some albums, and then faded during the mid to late 1980's. But they are still well known here in the Seattle area as being -- along with Queensryche, Culprit, TKO, and Metal Church -- as some of the first hard hitting rock bands to come out of this region, and to get known nationally, previous to the Grunge era of the 1990s. 

For those few Seattle people who may read this blog, I tried to find "Rockin You, Rockin Us" or "You've Got A Lot To Live", but the results I got using the Blogger YT search came up with numerous, completely unrelated results. :-)

IN OTHER LIFE...
Lately my car has been laid up. It ran great, but would simply stall after a couple miles of driving, and it wouldn't fire back up. It turned out it was several problems with the fuel system -- the gas tank filter was clogged, the gas line -- being a 1989 item -- was leaky, the dual fuel pumps were faulty as well. And, naturally, thanks to the after effects of the pandemic, the repair work was waiting on parts.

So, for the past two months I have learned to live without a car. Which means I have had to learn to shop for food online. Surprise! It actually works. And I also have learned to stretch out my food bill, and I do it by making vegetable stews and soups. Potatoes, carrots, onions, celery, and a can of beans, or some chili boiled together in a saucepan can make a decent meal that lasts 2 or 3 days. Add a dollop from a can of beef stew and it tastes even better.

The weather here is still a bit cold, but we haven't had a deep freeze in a couple weeks... Just one or two nights where it was icy or frosty out. One light dusting of snow (maybe half an inch at most -- it melted the next day), and it got down to 22F one night (-6C). That night apparently set a record low for that day in March.

I'm looking forwards to Spring and Summer, when we can see the Sun again.

In my MW/SW radio hobby, I've still been hearing the same 200-300 stations at night, and DX conditions have been fair to middling. The high SW bands have opened up a bit, but the activity levels aren't as high as they were 11 years ago during the same part of the Solar Cycle.

But, it's better listening than it was 5-6 years ago, when many nights it was total crap conditions on MW and horrible on SW, so I really have no reason to complain. You get the ionospheric conditions you get, after all. It's sort of like the weather. You can't control it. You learn to live with it.

So I do. :-)

SOME GRUNGE, AND SOME HENDRIX, AND...
I shall end this with a vid of the band Sweetwater, which I posted pics up earlier in this article. Here is Head Down, my favorite track by them. It sounded just as good at the Croc when I took the pics I posted.:

Keepin' my head down..... Keepin' my head down... Sweetwater's hit album track from 1993. Except it wasn't a hit on the radio. Should have been, though. They played it at the Croc. This is the official vid, probably from MTV. Remember them? When they played music? That's right, another rock tradition long gone. Anyway, enjoy this bit of Seattle music history. The Stratocaster player performs an awesome solo. 

For those who aren't Jimi Hendrix nuts, Randy Hansen has been doing accurate portrayals of Hendrix's playing and sound for years. He is a Seattle native, and is quite popular in Europe -- more popular there than he is here. Here he is in Austria, playing Who Knows, a bluesy track Hendrix did with the Band Of Gypsys. This song is one of my favorite Hendrix blues riffs, and Randy does a great job of it here in Vienna in 2012.

A pro shot taken in Vienna in 2012 for Randy Hansen's DVD that year. Hansen is more popular in places like Austria and Germany than he is in the US. I've seen him twice, the first time at the Algona Tavern in South King County, the second time in Kent when I took the pic I posted earlier in this blog article. He's well worth checking out. Even does versions of other late 60's hits the way Jimi Hendrix would have performed them. I heard him do a great job of Crosby Stills Nash & Young's "Almost Cut My Hair".

Well, that about sums things up for this time around. I hope you all got something out of the article; it was just something that struck me while I was looking at some AI generated pics that looked so much like the real thing that it made me wonder how it would affect one of my former favorite hobbies.

I don't think AI will affect guitar or radio DXing much... It may replace guitar players with AI players, but there aren't that many guitar players on new recordings anymore, not compared to the 1980's or 1990's, anyway. And radio will fully embrace AI, completely eliminating DJ's and perhaps most announcers -- despite some experts who claim otherwise -- but AI can't affect DXing. 

So at least those two hobbies are safe. :-)

Peace,
C.C., March 7th, 2012 (woops. You can see where my head was at when I finished writing and editing.... back in time I guess.... It's actually 2024!)














No comments:

Post a Comment